a small dropdown or spinbox would do. After we implement drag and drop of dimensions for transpose (#13), we could have a \ between row dimensions and columns dimensions and one could add/remove column dimensions by dragging them to either side of the .

I guess the spinbox is more appropriate for a number, and if we move filters to the headers (like I would like to), this will still work. I do not see how we could make the \ option work in that case.

There still might be a way to implement this using drag-and-drop in the "header-filter" case, if we switch from:

 a   b   c
        c0  c1  c2
a0  b0   0   1   2
a0  b1   3   4   5
a1  b0   6   7   8
a1  b1   9  10  11

to

 a   b
     c  c0  c1  c2
a0  b0   0   1   2
a0  b1   3   4   5
a1  b0   6   7   8
a1  b1   9  10  11

we could drag a row dimension (eg "a") on the "c" row to move it to columns:

 b                     
 a  a0  a0  a0  a1  a1  a1        
 c  c0  c1  c2  c0  c1  c2
b0   0   1   2   6   7   8
b1   3   4   5   9  10  11

Or, we could switch to Pandas representation (ie column headers before row dimension names), which seems more readable in this case but which does not reflect the axis order easily (b, a, c in this case).

a  a0       a1        
c  c0 c1 c2 c0  c1  c2
b                     
b0  0  1  2  6   7   8
b1  3  4  5  9  10  11